A Comprehensive Review of Drone-Based Autonomous Driving Datasets: Methodology, Taxonomy, and Prospects

Autonomous driving datasets (ADDs) have long served as a critical foundation for the development of autonomous driving technologies. Among them, drone-based ADD (DADD) refers to traffic flow trajectory datasets captured by drones. This paper reviews 17 currently available open-source DADDs and introduces evaluation metrics to assess their quality. By comparing and analyzing these datasets and related papers, we found that the elaboration of the dataset generation chain is incomplete, and we propose a generic generation framework for DADD, including four parts: data collection, trajectory acquisition, map construction, and traffic signal data acquisition. Specific optional methods are specified for each part, and optimization directions are proposed. In addition, we summarize the applications of DADD in the field of autonomous driving and intelligent transportation and deeply analyze the future application trends of DADD. Our work provides researchers not only with a guide for selecting open-source DADDs and generating DADDs but also with a guideline for the future construction and application of DADDs.